The Data Management Platform (DMP) is over: what replaced it

A Data Management Platform collected cookie IDs and third-party segments to build advertising audiences from. You placed a pixel, the platform recognised visitors through a third-party cookie, and you bought in segments other companies had assembled.

What came out was an anonymous audience you pushed to a DSP or an ad platform: "men, likely in market for a car", or "visitors who browsed travel sites". Names and email addresses usually stayed out of it, because the whole thing ran on cookie IDs.

That model held as long as one cookie could follow a visitor across different websites. That is exactly the part that disappeared.

Browsers. Safari has blocked third-party cookies by default for years, and so has Firefox. Chrome announced the end, postponed it repeatedly, and eventually went a different way. It hardly matters: once a large share of your reach accepts no third-party cookie, an audience built on them cannot be trusted.

Regulation. Under GDPR and the ePrivacy rules you need consent before tracking someone for advertising, plus a legal basis for passing that data on. European regulators have taken a hard look at the ad industry's consent framework. Buying audiences turned into a legal project rather than a purchase order.

Quality. The segments were mediocre anyway. Inferred from browsing behaviour, often at household level, sometimes months old, and nobody could explain how a given profile was built.

The market moved accordingly. Oracle wound down its advertising business, and the category as a whole appears on few shortlists now.

A Customer Data Platform runs on first-party data: behaviour on your own site and app, orders, email interaction, support contacts. It ties those events to a person, usually by email address or customer number, and builds a profile from them.

From that profile you cut audiences: customers who ordered twice and have been quiet for ninety days, or trial users who never touched the core feature. Those lists go back out to your email platform and your ad platforms.

The difference with a DMP is provenance. You use data from people who have a relationship with you, and you know where every attribute came from. That holds up in front of a regulator, and it performs better.

In plenty of cases the underlying model already exists. If your orders, web behaviour and email stats sit in BigQuery or Snowflake, an audience is a query:

SELECT customer_id, email_sha256
FROM `my-project.customers.profile`
WHERE order_count >= 2
  AND last_order_date < DATE_SUB(CURRENT_DATE(), INTERVAL 90 DAY)
  AND marketing_consent = TRUE

What you still need is an activation layer that writes the result out to Google Ads and Meta. That is reverse ETL, with tools like Census or Hightouch, and at heart it is a scheduled sync.

Both platforms accept hashed email addresses, through Customer Match and Custom Audiences. Expect losses on matching, because the address in your database is not always the one someone signed up with there, and both platforms require a minimum list size before they will serve an audience.

A CDP earns its licence when you recognise the following:

• Many channels that all need the same audiences. With two channels, an export is built faster than a CDP is configured.
• Marketers who want to assemble audiences themselves, without SQL and without a ticket to the data team. That is the strongest reason, and the one a warehouse does not solve.
• A genuine realtime need: a response within seconds of on-site behaviour. An hourly sync covers most newsletter segmentation with room to spare.
• Identity resolution across channels that is messy enough that you would rather not maintain it yourself.

A warehouse with an export layer is enough when your data already lives there, audiences may refresh in hours or days, and only a handful of people build audiences. That saves you a licence, an implementation project, and a second place where the definition of "active customer" lives.

• Still running bought third-party segments? Hold them against your own conversion data. Often there is no visible difference with the campaign that ran without them.
• Get your first-party data in order before picking a platform. A CDP on top of messy data produces faster mess.
• Check what your warehouse already does. If the audience is a query, you need a sync, not a new platform.
• Record consent per purpose and carry it into every audience query. A list without a consent field is not a list you may export.
• Hash email addresses with SHA-256 before they leave your systems, and track your match rate per platform.
• Pick a CDP when marketers need to build audiences themselves. If you are picking one for a realtime promise, first name the use case that needs those seconds.